Extended Calculation of Dark Matter-Electron Scattering in Crystal Targets

CALT-TH-2021-018
Extended Calculation of Dark Matter-Electron

Scattering in Crystal Targets
arXiv:2105.05253v1 [hep-ph] 11 May 2021
Sinéad M. Griffin,a Katherine Inzani,a Tanner Trickle,b Zhengkang Zhangb and

Kathryn M. Zurekb
a
Material Sciences Division and Molecular Foundry,
Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
b
Walter Burke Institute for Theoretical Physics,
California Institute of Technology, Pasadena, CA 91125, USA
Abstract: We extend the calculation of dark matter direct detection rates via electronic
transitions in general dielectric crystal targets, combining state-of-the-art density func-
tional theory calculations of electronic band structures and wave functions near the band
gap, with semi-analytic approximations to include additional states farther away from the
band gap. We show, in particular, the importance of all-electron reconstruction for recov-
ering large momentum components of electronic wave functions, which, together with the
inclusion of additional states, has a significant impact on direct detection rates, especially
for heavy mediator models and at O(10 eV) and higher energy depositions. Applying our
framework to silicon and germanium (that have been established already as sensitive dark
matter detectors), we find that our extended calculations can appreciably change the de-
tection prospects. Our calculational framework is implemented in an open-source program
EXCEED-DM (EXtended Calculation of Electronic Excitations for Direct detection of Dark
Matter), to be released in an upcoming publication.
Contents
1 Introduction 1
1.1 Overview of the Calculation and Key Results 3
2 Electronic States 5
2.1 DFT Wave Functions and Band Structures 6
2.1.1 All-electron Reconstruction 8
2.2 Atomic Wave Functions 9
2.3 Plane Wave Approximation 11
3 Electronic Transition Rates 12

3.1 Valence to Conduction 14
3.2 Valence to Free 16
3.3 Core to Conduction 17
3.4 Core to Free 19
3.5 In-medium Screening 21
4 Projected Sensitivity 22
4.1 Comparison With Previous Results 25
5 Conclusions 28
1 Introduction
Electronic excitations have been established as an alternative to nuclear recoils in direct

detection of sub-GeV dark matter (DM). Nuclear recoil searches lose sensitivity at lower
DM masses due to kinematic mismatch between the DM and heavier nuclei, whereas elec-
tronic transitions can potentially extract all of the DM kinetic energy during a DM-electron
scattering event by excitation across an energy gap. Proposed targets, including noble gas
atoms with O(10 eV) ionization energies [1–8], semiconductors with O(eV) electronic band
gaps [1–3, 9–18], and superconductors and Dirac materials with O(meV) band gaps [19–25],
extend the reach on DM mass well below the limit of nuclear recoil. Experimental searches
using dielectric crystal targets are currently underway, specifically with Si (DAMIC [26–
28], SENSEI [29–32], SuperCDMS [33–39]) and Ge (EDELWEISS [40–42], as well as Su-
perCDMS) which have been predicted to have excellent sensitivity down to O(MeV) DM
masses based on their O(eV) band gaps.
Reliable theoretical predictions of target-specific transition rates are important not only
for current experiments, but also for planning the next generation of detectors. Compared
to the DM-induced electron ionization rate in noble gases like xenon [1–8], calculations for
–1–
the DM-electron scattering rate in a crystal are more complicated. Ionization rates for
noble gases can be calculated by considering each noble gas atom as an individual target,
where the calculation simplifies to finding the ionization rate from an isolated atom, for
which the wave functions and energy levels are well tabulated [43]. However, for crystal
targets the atoms are not isolated and more involved techniques are required to obtain an
accurate characterization of DM-electron interactions in a many-body system.
There have been a variety of approaches taken to compute the DM-electron scattering
rate in crystals. One of the first attempts, Ref. [2], computed the rate with semi-analytic
approximations for the initial and final state wave functions, and used the density of states
to incorporate the electronic band structure. Later, Ref. [3] continued in this direction and
used improved semi-analytic approximations for the initial state wave functions. Meanwhile,
a fully numerical approach was advanced in Refs. [1, 10, 11] where density functional theory
(DFT) was employed to calculate the valence and conduction electronic band structures
and wave functions. The latter approach, as implemented in the QEdark program and
embedded in the Quantum ESPRESSO package [44–46], has become the standard for
first-principles calculations of DM detection rates. Recently, in Refs. [15, 16] we used a
similar DFT approach as implemented in our own program for a study of DM-electron
scattering in a variety of target materials. More recently there has been work utilizing the
relation between the dielectric function and the spin-independent scattering rate [47–49],
which also properly incorporates screening effects.
The goal of this work is to further extend the DM-electron scattering calculation in
several key aspects, and present state-of-the-art predictions for Si and Ge detectors using
a combination of DFT and semi-analytic methods. As we will elaborate on shortly, the
time- and resource-consuming nature of DFT calculations presents an intrinsic difficulty
that has limited the scope of previous work in this direction to a restricted region of phase
space; typically only bands within a few tens of eV above and below the band gap were
included and electronic wave functions were cut off at a finite momentum. We overcome
this difficulty by implementing all-electron (AE) reconstruction (whose importance was
previously emphasized in Ref. [50]) to recover higher momentum components of DFT-
computed wave functions, and by extending the calculation to bands farther away from
the band gap using semi-analytic approximations along the lines of Refs. [2, 3]. As we will
see, the new contributions computed here have a significant impact on detection prospects
in cases where higher energy and/or momentum regions of phase space dominate the rate,
including scattering via a heavy mediator, and experiments with O(10 eV) or higher energy
thresholds. We also stress that in contrast to the recent work emphasizing the relation
between spin-independent DM-electron scattering rates and the dielectric function [47–49],
our calculation can be straightforwardly extended to DM models beyond the standard spin-
independent coupling. Furthermore, we do not make assumptions about isotropy for the
majority of our calculation, and our framework is capable of treating anisotropic targets
which exhibit smoking-gun daily modulation signatures [15, 23, 24] (see also Refs. [51, 52]
for discussions of daily modulation for phonon excitations).
Our calculation is implemented in an open-source program EXCEED-DM (EXtended Cal-
culation of Electronic Excitations for Direct detection of Dark Matter), to be released in
–2–
Si Ge
free
free
60 eV (Edft) 60 eV (Edft)
val cond
val cond
1.11 eV (Eg ) 0.67 eV (Eg )
0 eV 0 eV
−12 eV −14 eV
−28 eV (3d)
−116 eV (2p) −140 eV (3p)
core
core
−170 eV (2s) −195 eV (3s)
−1.3 keV (2p)
−1.4 keV (2s)
−1.9 keV (1s) −11 keV (1s)
Figure 1. Schematic representation of electronic states in Si (left) and Ge (right), divided into
core, valence (“val”), conduction (“cond”) and free. Shaded regions indicate the range of energies
for each type of electronic states. In a scattering process, electrons transition from either core or
valence states, below the Fermi surface at E = 0, to conduction or free states above the band
gap Eg . As outlined in Sec. 1.1 and explained in detail in Sec. 2, we compute the valence and
conduction states numerically using DFT (including all-electron reconstruction), model the core
states semi-analytically with RHF wave functions, and treat the free states as plane waves.
an upcoming publication. Currently a beta version of the program is available here [53]. We
also make available our DFT-computed wave functions [54] and the output of EXCEED-DM [55]
for Si and Ge.
1.1 Overview of the Calculation and Key Results

Before delving into the technical details, let us give a brief overview of the calculation and
highlight some key results. We divide the electronic states in a (pure) crystal into four
categories: core, valence, conduction and free, as illustrated in Fig. 1 for Si and Ge and
discussed in more detail in Sec. 2. At zero temperature, electrons occupy states up to
the Fermi energy, defined as the maximum of the valence bands and denoted by E = 0.
The band gap Eg , i.e. the energy gap between the occupied valence bands and unoccupied
conduction bands, is typically O(eV) for semiconductors, e.g. 1.11 eV for Si and 0.67 eV for
Ge; this sets a lower limit on the energy deposition needed for an electron transition to
happen.
The electronic states near the band gap deviate significantly from atomic orbitals and
need to be computed numerically. We apply DFT methods (including AE reconstruction)
for this calculation, and refer to the DFT-computed states as valence and conduction.
Specifically, for both Si and Ge, we take the first four bands below the gap to be valence,
which span an energy range of −12 eV to 0 and −14 eV to 0, respectively, and take bands
above the gap up to Edft = 60 eV to be conduction.
–3–
1 10−37
QEdark [11]
10−1 10−38
5
∆Rω × kg · yr
≥
Q
1
≥
10−2 c→c 10−39
σ e [cm2]
Q
],
[10
v→c
5
≥
rk
Q
da
1
10−3 10−40
QE
≥
Q≥
Q
v→f
rk,
5
wo
is
10−4 10−41
Th
c→f
Lee et al. [3], Q ≥ 1
−1 2 3
0 25 50 75 100 125 150 175 200 10 1 10 10 10 104
ω [eV] mχ [MeV]
Figure 2. Selection of results from Sec. 4, for DM-electron scattering via a heavy mediator in a
Ge target. Left: Contribution from each of the four transition types, valence to conduction (v→c),
valence to free (v→f), core to conduction (c→c), and core to free (c→f) to the scattering rate
binned in energy deposition (with ∆ω = 1 eV) for a 1 GeV DM at a given reference cross section
σ e = 10−40 cm2 . Right: 95% C.L. projected limit (3 events) on σ e assuming 1 kg-year exposure,
for energy thresholds corresponding to 1 and 5 electron-hole pairs. We compare our results with
QEdark calculations in Refs. [10, 11] and the semi-analytic model of Lee et al [3]; see text for details.
With more computing power we can in principle include more states, both below and
above the band gap, in the DFT calculation. However, since the states far from the band
gap can be modeled semi-analytically with reasonable accuracy, computing them with DFT
is inefficient. Below the valence bands, electrons are tightly bound to the atomic nuclei.
We model them using semi-analytic atomic wave functions and refer to them as core states.
These include the 1s, 2s, 2p states in Si and 1s, 2s, 2p, 3s, 3p, 3d states in Ge (the 3d
states in Ge are sometimes referred to as semi-core, and we will compare the DFT and
semi-analytic treatment of them in Secs. 2.2 and 3.3). Finally, above Edft = 60 eV, we
model the states as free electrons as they are less perturbed by the crystal environment.
With the electronic states modeled this way, we compute the rate for valence to con-
duction (v→c), valence to free (v→f), core to conduction (c→c) and core to free (c→f)
transitions induced by DM scattering, as discussed in detail in Sec. 3. The total rate is the
sum of all four contributions. We then use our calculation to update the projected reach of
direct detection experiments in Sec. 4, and compare our results with previous literature.
Figure 2 gives a glimpse of some of our key results. Here we consider the case of DM
scattering via a heavy mediator in a Ge target. The impact of core (3d) to conduction
contributions is clearly visible from both the differential rate (left panel, for mχ = 1 GeV)
and the projected reach (right panel). They dominate the total rate for mχ & 10 MeV,
and, as we can see from the right panel of Fig. 2, lead to significantly more optimistic reach
compared to previous DFT calculations implemented in QEdark [10, 11]; this is especially
–4–
true for higher detector thresholds (corresponding to higher Q values). Note that while
Refs. [10, 11] included the 3d states in their DFT calculation, their contributions were
significantly underestimated due to the absence of AE reconstruction. The importance of
AE reconstruction is also seen from the valence to conduction differential rate in the left
panel of Fig. 2, where our calculation predicts a much higher rate at ω & 15 eV compared
to the QEdark calculation in Ref. [11]. Meanwhile, accounting for in-medium screening (see
Sec. 3.5) we find, consistent with Ref. [48], a lower rate at energy depositions just above
the band gap, and as a result, weaker reach at low mχ , compared to Refs. [10, 11]. On the
other hand, our modeling of the core (3d) states is similar to the semi-analytic approach
of Ref. [3], and indeed we find very similar reach at large mχ ; however, the approach of
Ref. [3] overestimates the rate at smaller mχ due to reduced accuracy in the modeling of the
valence and conduction states. We reserve a more detailed comparison with the literature
for Sec. 4.1.
2 Electronic States
To compute the DM-electron scattering rate one must understand the electronic states of
the target. In targets with a periodic potential, Bloch’s theorem states that the energy
eigenstates can be indexed by a momentum, k, which lies within the first Brillouin zone
(1BZ). These Bloch states, ψi,k , where i represents additional quantum numbers, are eigen-
states of the discrete translation operator such that ψi,k (x + r) = eik·r ψi,k (x), which means
the electronic wave functions can be written as
1
ψi,k (x) = √ eik·x ui,k (x) , (2.1)
V
where ui,k (x + r) = ui,k (x) and V is the target volume. For every k there exists a tower of
eigenstates (labeled by i) of the target Hamiltonian which constitutes the complete set of
states in the target. Unfortunately this complete set is not known for a general material and
therefore a combination of approximations must be used to calculate them. As discussed
in Sec. 1.1 and illustrated in Fig. 1, we divide the states into core, valence, conduction and
free, and use a combination of numerical calculations and semi-analytic modeling. In this
section, we expand on the treatment of each type of electronic states.
We first discuss the DFT calculation for valence and conduction states in Sec. 2.1, and
then move on to explain the semi-analytic treatment of core states in Sec. 2.2. Our main
results are contained in Fig. 3 where we compare the average magnitude of electronic wave
functions, binned in momentum (see Eq. (2.7)), computed with and without AE recon-
struction, discussed further in Sec. 2.1.1, and, for the highest energy core states (2p in Si
and 3d in Ge), those computed using the core approximation discussed in Sec. 2.2. We find
that the AE reconstruction includes a significant contribution from wave functions at large
momentum as expected, and that for the core states, the semi-analytic approach reproduces
the large momentum components of these AE reconstructed DFT wave functions. Lastly
we will discuss the analytic treatment of the free states in Sec. 2.3.
–5–
1 1
−1
Si − val 10 Si − 2p
10−1 10−2
10−2 10−3 AE
10−4 no AE

ui | 2
ui | 2
10−3 10−5 core
|e
|e
10−6
10−4 10−7
10−5 10−8
10−9
0 10 20 30 40 0 10 20 30 40 50 60 70 80
q [keV] q [keV]
1 10−2
Ge − val 10−3 Ge − 3d
10−1
10−4
10−2
10−5

ui | 2
ui | 2
−3
10
|e
|e
10−6
10−4
10−7
10−5 10−8
0 10 20 30 40 0 10 20 30 40 50 60 70 80
q [keV] q [keV]
Figure 3. Comparison of the Bloch wave function magnitudes, defined in Eq. (2.7), computed
with DFT with (red, “AE”) and without (blue, “no AE”) AE reconstruction, and the semi-analytic
core approximation of Eq. (2.9) (green, “core”). Shaded bands indicate the maximum and minimum
values across all the bands belonging to the state type indicated in the upper right corner of each
panel. AE reconstruction, discussed in Sec. 2.1.1, recovers the large momentum behavior of the
electronic wave functions. Core electronic states, such as those shown in the right panels and
discussed in Sec. 2.2, can be well modeled semi-analytically with atomic wave functions, as seen
by the good agreement between the “core” and “AE” curves. When applicable, the semi-analytic
parameterization is advantageous since the electronic wave functions are then known to arbitrarily
large momentum.
2.1 DFT Wave Functions and Band Structures
In principle, DFT provides an exact solution to the many-electron Schrödinger equation by

the Hohenberg-Kohn theorems that treat all properties of a quantum many-body system
as unique functionals of the ground state density. They further show that the exact ground
–6–
Si Ge
12
8 10
4 5
Energy [eV]
Energy [eV]
0 0
−4 −5
−8 −10
−12
Γ X U|K Γ L W X Γ X U|K Γ L W X
Figure 4. Calculated band structures of Si (left) using a PBE xc-functional within DFT and
Ge (right) using a hybrid functional HSE06. The band gaps have been scissor corrected to their
measured values near zero temperature, 1.11 eV and 0.67 eV for Si and Ge, respectively. The Fermi
level is set to 0 eV in both panels.
state density and energy can be found by minimizing the total energy of the system [56, 57].
This becomes tractable by the Kohn-Sham (KS) equations that reduce the many-body
problem to non-interacting electrons moving in an effective potential, Veff ,

p2
+ Veff − i ψi = 0 , Veff = Vext + VH + Vxc , (2.2)
2me
where i is the orbital energy of the KS orbital ψi [58]. The external potential Vext and
Hartree potential VH , are known, while the exchange-correlation (xc) potential Vxc , which
contains the many-body interactions, must be approximated. Herein lies the deviation
from the exact solution, and although various formulations of xc-energy functionals have
been successful, the choice of xc-functional will affect the predicted electronic states and
hence calculated transition rates. For Si, we use PBE [59], a type of generalized gradient
approximation (GGA) xc-functional which is one of the most popular and low-cost choices.
Local and semi-local based xc-functionals, such as PBE, suffer from a self-interaction er-
ror and band gap underestimation, which we modify with a “scissor correction” where the
bands are shifted to match the experimentally determined values of band gap. For Ge,
this underestimation results in zero band gap with PBE, therefore we instead use a hybrid
functional, which mixes a parameterized amount of exact exchange into the xc-functional,
correcting band gaps and band widths by error cancellation at the cost of increased com-
putation time. We use the range-separated hybrid functional HSE06 [60, 61], which applies
a screened Coulomb potential to correct the long-range behavior of the xc-potential, giving
high accuracy at a mid-level computational cost. Our computed band structures for Si and
Ge are shown in Fig. 4.
The periodic Bloch wave functions, ui,k (x), Eq. (2.1), for band i and Bloch wave vector
k are computed by finding the Fourier coefficients, uei,k,G (which satisfy the normalization
–7–
P 2
condition, G |e
ui,k,G | = 1):
X
ui,k (x) = ei,k,G eiG·x .
u (2.3)
G
The number of reciprocal lattice vectors G kept in the sum is conventionally set by an energy
cutoff, Ecut , such that |k+G|2 < 2me Ecut . These Bloch wave function coefficients u
ei,k,G for
both Si and Ge are computed with the projector augmented wave (PAW) method [62, 63]
within vasp [64–67] up to Ecut = 1 keV on a 10 × 10 × 10 uniform k mesh over the 1BZ. We
then include AE reconstruction effects up to a higher energy cutoff, EAE = 2 keV, which
recovers higher momentum components of the wave functions up to |k + G|2 < 2me EAE ,
as discussed in more detail in Sec. 2.1.1. The Bloch wave function coefficients, u ei,k,G , for
Si and Ge used for this work can be found here [54].
A final consideration of using DFT wave functions is that DFT is fundamentally a
ground state method, and the KS conduction band states are only approximations to excited
states. Excited state methodologies are much more computationally expensive than ground
state KS-DFT. Furthermore, since excited state quasiparticles, such as excitons, have been
argued to have a negligible effect on the calculation of DM scattering rates [11], they are
neglected in our calculations.
2.1.1 All-electron Reconstruction

There are many different approaches to find the eigenstates of Eq. (2.2). The PAW
method [62] is one such standard approach. The main idea of the PAW method is to split
up the calculation of the eigenstates: near the ionic centers the wave functions resemble
the eigenstates of an isolated atom, while further away they can be computed numerically
with a pseudopotential. This greatly simplifies the numeric calculation since the focus is
then on large distance (small momentum), and the small distance (high momentum) pieces
can be self-consistently reintroduced after the main part of the DFT calculation. The large
distance components of the wave function are known as “pseudo wave functions” (PS wave
functions) and the total wave functions are known as the “all-electron wave functions” (AE
wave functions), indicating that all of the wave function components are included. We
will now give a brief overview of how the AE wave functions can be reconstructed from
the PS wave functions, computed with PAW-based DFT codes, and refer the reader to
Refs. [62, 63, 68, 69] for more detailed information.1
The AE wave functions, |ΨAE i are built from two components. Near the ionic core,
or inside an “augmentation sphere,” |ΨAE i is expanded in a set of basis functions, |φAE
j i,
which are simply taken to be the wave functions of an isolated atom,
X
|ΨAE i = cj |φAE
j i. (2.4)
j
1
It is possible to calculate all electronic eigenstates, including the core, self-consistently by other more
complex methods such as the full-potential linearized augmented plane wave (FP-LAPW) method or the
relaxed-core PAW (RC-PAW) method.
–8–
Outside of the augmentation sphere, |ΨAE i = |ΨPS i. Near the ionic core the PS wave
functions |ΨPS i are expanded in a set of basis functions |φPS
j i that are computationally
AE
more convenient than the |φj i. Therefore,
X X
|ΨAE i = |ΨPS i − c0j |φPS
j i+ cj |φAE
j i, (2.5)
j j
which simply replaces the components in |ΨPS i within the augmentation sphere with the
P
AE wave function. To find the c coefficients we insert an identity, 1 = j |φAE AE
j ihpj | =
P PS PS AE/PS
j |φj ihpj |, where |pj i are projector functions, defined to satisfy this identity within
the augmentation sphere. Therefore, cj = hpAE AE i, c0 = hpPS |ΨPS i. The last ingredient
j |Ψ j j
to compute |Ψ i from |Ψ i is to require that |φAE
AE PS
j i is related to |φPS
j i via a transforma-
AE PS
tion, |φ i = T |φ i. This implies that all the PS states are related to AE states by this
transformation T , such that cj = c0j and the AE reconstruction can be written as
X PS PS
|ΨAE i = |ΨPS i + |φAE PS
j i − |φj i hpj |Ψ i . (2.6)
j
In practice, we implement the AE reconstruction with pawpyseed [69], and the plane wave
expansion cutoff of |ΨAE i, EAE , can be increased from the initial Ecut . We use EAE = 2
keV.
To visualize the effect of AE reconstruction, we plot in Fig. 3 the average magnitude
of the Bloch wave functions, binned in q,
D E 1 XX
ui |2 (q; ∆q) ≡
|e ui,k,G |2 θ(q + ∆q − |k + G|) θ(|k + G| − q) ,
|e (2.7)
Nq
k G
where uei,k,G are the Fourier components of the Bloch wave functions, defined in Eq. (2.3).
Each bin in momentum space extends from q to q + ∆q with ∆q = 1 keV, and Nq is a
P P
normalization factor equal to the number of points in a bin, Nq = k G θ(q + ∆q −
|k + G|) θ(|k + G| − q). We see that AE reconstruction recovers the high momentum
components, which as we will see can significantly affect the DM-induced transition rate
for processes which favor large momentum transfers (such as processes mediated by heavy
particles), or processes limited to larger ω (e.g. higher experimental thresholds where large
q processes are the only kinematically allowed transitions). Previous DFT calculations of
DM-induced electron transition rates, with the exceptions of Ref. [15, 16, 50], used only
the pseudo wave functions, |ΨPS i as opposed to the AE wave functions, |ΨAE i, and have
therefore underestimated detection rates in several cases.
2.2 Atomic Wave Functions

If one could reconstruct the AE wave functions arbitrarily deep into the band structure,
and to arbitrarily high momentum, one could calculate an accurate representation of the
complete set of electronic states with a DFT calculation. In practice, however, this is
neither feasible nor necessary. States deep in the band structure are more isolated from
the influence of the crystal environment, and so an isolated atomic approximation becomes
–9–
valid. We refer to these inner, tightly bound electrons as core electrons. In Si, we will show
that the 2p states and below can be treated as core, while in Ge, the 3d states and below
can, as alluded to in Fig. 1. The purpose of this subsection is to expand on the atomic
approximation for core electrons and discuss its accuracy.
More precisely, the initial states of a transition should be taken as a linear combination
of isolated atomic wave functions that is in Bloch form (known as Wannier states):
1 X ik·(r+xκ ) atom
ψκnlm,k (x) = √ e ψκnlm (x − r − xκ ) , (2.8)
N r
where κ labels the atom in the primitive cell, n, l, m are the standard atomic quantum
P
numbers, xκ is the equilibrium position of the κth atom, r sums over all primitive cells in
the lattice, and N is the total number of cells. In contrast to the valence and conduction
states discussed in the previous subsection, the core states are labeled by (κnlm) rather
than band index i. The corresponding periodic (dimensionless) u functions can be easily
obtained via Eq. (2.1):
√ X −ik·(x−r−x ) atom
uκnlm,k (x) = Ω e κ
ψκnlm (x − r − xκ ) , (2.9)
r
where Ω = V /N is the primitive cell volume.

atom are not known analytically, but are ex-
In general, the atomic wave functions ψκnlm
panded in a basis of well-motivated analytic functions. The basis coefficients are then fit by
solving the isolated atomic Hamiltonian, giving a semi-analytic expression for ψκnlmatom . We
use a basis of Slater type orbital (STO) wave functions whose radial component is
n+ 21 n−1
−3/2 (2Z) r
RSTO (r; Z, n) = a0 p e−Zr/a0 , (2.10)
(2n)! a0
where a0 = 0.53 Å = (3.7 keV)−1 is the Bohr radius, and Z is an effective charge of the ionic
potential. Including the angular part, the atomic wave functions are then
X
atom
ψκnlm (x) = Cjln,κ RSTO (x; Zjl,κ , njl,κ )Ylm (x̂) , (2.11)
j
where Cjln,κ , Zjl,κ , njl,κ are tabulated in Ref. [43], and Ylm (x̂) are the spherical harmonics
with the Condon-Shortley phase convention [70].
To assess the accuracy of the atomic wave function approximation, we temporarily push
the DFT calculation beyond its default regime (valence and conduction), to the highest core
states – 2p states in Si and 3d states in Ge, where it is still computationally feasible – and
compare the numerical wave functions to the semi-analytic ones discussed above. The
results, in terms of the average magnitude of Bloch wave functions defined in Eq. (2.7),
are shown in the right panels of Fig. 3.2 We see that the atomic approximation accurately
2
The flatness of band structures offers a complementary check of the validity of the atomic approximation.
We have verified that the DFT computed energy eigenvalues indeed have a small variance for the highest
core states, as expected.
– 10 –
√
reproduces the numerical wave functions up to the momentum cutoff 2me EAE ' 50 keV
for EAE = 2 keV. These plots also show the limitation of DFT calculations. While AE
reconstruction recovers higher-momentum components of electronic wave functions, it is
not feasible to expand the plane wave basis set to arbitrarily high cutoff. However, having
verified the atomic approximation for the highest core states, we can use it for all core
states with confidence, allowing us to more easily include the high momentum components
beyond the DFT cutoff.
2.3 Plane Wave Approximation
With the inclusion of the semi-analytic core states, all of the states below the band gap
have been modeled. States above the band gap can also be computed with DFT methods,
as described in Sec. 2.1. Similar to valence bands, there are practical limitations to how
many conduction bands can be included. To remedy this in the simplest way possible, we
model states far above the band gap as plane waves,
1 |k + G|2
ψG,k (x) = √ ei(k+G)·x , EG,k = , (2.12)
V 2me
where G is a reciprocal lattice vector, and plays the role of a band index. (To understand
this, simply note that every momentum can be decomposed into a k vector inside the
1BZ and a reciprocal lattice vector. Integrating over the momentum of plane wave states
amounts to a k integral within the 1BZ and a G sum.) From Eq. (2.1) we see that the
corresponding periodic u functions are simply
uG,k (x) = eiG·x . (2.13)
The plane wave approximation is often used in atomic ionization rate calculations, with the
inclusion of a Fermi factor, F (ν),
ν αme
F (ν) = , ν(Zeff , E) = 2πZeff √ , (2.14)
1 − e−ν 2me E
where E is the final state electron energy, and Zeff is an effective charge parameter, which
enhances the transition rate at low E to account for the long range behavior of the Coulomb
potential. See Refs. [1–3, 6] for more details. In atomic ionization calculations one usually
takes Zeff to be related to the binding energy of the initial state, EB ,
r
EB
Zeff = n , (2.15)
13.6 eV
where n is the principal quantum number. Since the rate is proportional to the Fermi
factor, Zeff = 1 is seen as the conservative choice. Later in Secs. 3.2 and 3.4 we quantify
how much of an effect this has on the transition rate. This uncertainty is only important for
very high experimental thresholds, and generally we find that Zeff = 1 leads to a smoother
match (within an O(1) factor) to conduction band contributions from DFT calculations.
– 11 –
3 Electronic Transition Rates
We now present the DM-induced electron transition rate calculation. We begin with a
general discussion and then in Secs. 3.1-3.4 consider the four different transition types in
turn: valence to conduction (v → c), valence to free (v → f), core to conduction (c → c) and
core to free (c → f). Finally, in Sec. 3.5 we discuss the treatment of in-medium screening.
The general derivation has been discussed previously (see e.g. Refs. [2, 5, 10, 15, 50]),
and we repeat it here for completeness and clarity, as a variety of conventions have been
used. Beginning with Fermi’s Golden Rule, the transition rate between electronic states
|i, si and |f, s0 i due to scattering with an incoming non-relativistic DM particle, χ, with
mass mχ , velocity v, and spin σ is given by
Z
d3 q 0 0 2
Γi,s,σ→f,s0 ,σ0 (v) = 2πV 3
hp , σ ; f, s0 | δ Ĥ |p, σ; i, si δ(Ef,s0 − Ei,s − ωq ) , (3.1)
(2π)
where |p, σ; i, si = |p, σi ⊗ |i, si, q is the momentum deposited onto the target, p = mχ v,
p0 = p − q, δ Ĥ is the interaction Hamiltonian, V is total volume of the target, and ωq is
the energy deposition:
1 (mχ v − q)2 q2
ωq = mχ v 2 − =q·v− . (3.2)
2 2mχ 2mχ
We assume that all quantum states are unit normalized. Modulo in-medium screening
effects, discussed below in Sec. 3.5, we can write Eq. (3.1) in terms of the standard QFT
matrix element, defined with plane wave incoming and outgoing states, by inserting 1 =
P R d3 k
V s (2π) 3 |k, sihk, s| and using
(2π)3 Mσ0 s0 σs (p0 , k0 , p, k) (3) 0

hp0 , σ 0 ; k0 , s0 | δ Ĥ |p, σ; k, si ≡ 2
δ p + k0 − p − k . (3.3)
V 4me mχ
We find
Z
2π d3 q
Γi,s,σ→f,s0 ,σ0 (v) = 2 2
δ(Ef,s0 − Ei,s − ωq )
16V me mχ (2π)3
Z 2
d3 k
× e e
Mσ0 s0 σs (p − q, k + q, p, k) ψf (k + q)ψi (k) ,
∗
(3.4)
(2π) 3
√
where ψei (k) = V hk|ii.
We will limit our analysis to matrix elements which only depend on q, and assume
that the electron energy levels are also spin independent, which allows the spin sums to be
easily computed:
1 XX
Γi→f ≡ Γi,s,σ→f,s0 ,σ0
2 0 0
σ,σ s,s
Z
4π d3 q
= |M(q)|2 |fi→f |2 δ (Ef − Ei − ωq ) , (3.5)
16V m2e m2χ (2π)3
Z Z
d3 k e∗ e
fi→f ≡ ψ (k + q) ψi (k) = d3 x eiq·x ψf∗ (x) ψi (x) , (3.6)
(2π)3 f
– 12 –
where |M|2 is the spin averaged matrix element squared and we have defined a crystal form
factor fi→f , written in terms of both momentum and position space representations of the
wave functions.
The transition rate per target mass, Ri→f , is then given by
Z
1 ρχ
Ri→f = d3 vfχ (v) Γi→f , (3.7)
ρT mχ
where ρT is the target density, ρχ = 0.4 GeV/cm3 is the local DM density, and fχ is taken
to be a boosted Maxwell-Boltzmann distribution. The total rate, R, is then simply the
sum over all possible transitions from initial to final states. Since the only v dependence in
Eq. (3.7) comes from the energy conserving delta function, we perform the v integral first
R
and define g(q, ω) = 2π d3 vfχ (v)δ(ω − ωq ). This integral can be evaluated analytically
(see e.g. Refs. [15, 23, 52]):
2π 2 v02 1 −v−2 /v 2 2
0 − e−vesc /v0
2

g(q, ω) = e , (3.8)
N0 q

1 q2

v− = min ω+ + q · ve , vesc , (3.9)
q 2mχ
where ω = Ef − Ei is the deposited energy, and N0 is a normalization factor such that
R 3
d vfχ (v) = 1. We take the DM velocity distribution parameters to be v0 = 230 km/s ,
vesc = 600 km/s, and ve = 240 km/s. The total rate then becomes
Z
2 ρχ X d3 q
R= |M(q)|2 g(q, ω) |fi→f (q)|2 . (3.10)
16V m2e m3χ ρT (2π)3
i,f
Here we focus on simple DM models, such as the kinetically mixed dark photon or lep-
tophilic scalar mediator models. In these models M(q) can be factorized as M(q) =

M(q0 )Fmed (q0 /q) fe /fe0 , where Fmed (q0 /q) = 1 for a heavy mediator and Fmed (q0 /q) =
(q0 /q)2 for a light mediator, and fe /fe0 is a screening factor discussed in more detail in
Sec. 3.5. As in previous works, we choose the reference momentum transfer to be q0 = αme .
We can then finally write the rate in terms of a reference cross section,
µ2χe
σe = |M(q0 )|2 , (3.11)
16πm2χ m2e
and find
Z 2
2πσ e ρχ X d3 q fe
R= 2 3 0
2
Fmed g(q, ω) |fi→f (q)|2 . (3.12)
V µχe mχ ρT (2π) fe
i,f
Another useful quantity is the binned rate (the rate for energy deposition between ω and
ω + ∆ω), ∆Rω , defined as
2πσ e ρχ X
∆Rω = θ(ω + ∆ω − Ef + Ei ) θ(ω − Ef + Ei )
V µ2χe mχ ρT
i,f
Z 2
d3 q fe
× 2
Fmed g(q, ω) |fi→f (q)|2 . (3.13)
(2π)3 fe0
– 13 –
1 10−1
10−1 − d log Fmed
d log q = 2 10−2 − d log Fmed
d log q = 2
10−2
∆Rω × kg · yr
∆Rω × kg · yr
10−3
−3
10
10−4
10−4
10−5
10−5
no AE AE
10−6 no AE AE
10−6
Si Ge
0 10 20 30 40 50 60 0 10 20 30 40 50 60
ω [eV] ω [eV]
10 10
Fmed Fmed
1 − d log
d log q = 0 1 − d log
d log q = 0
∆Rω × kg · yr
∆Rω × kg · yr
10−1 10−1
10−2 AE 10−2 AE
10−3 10−3
10−4 no AE 10−4 no AE
Si Ge
0 10 20 30 40 50 60 0 10 20 30 40 50 60
ω [eV] ω [eV]
Figure 5. DM-electron scattering rate from valence to conduction bands binned in energy deposi-
tion (with ∆ω = 1 eV) for 1 GeV DM, light (top row) and heavy (bottom row) mediators, assuming
σ e = 10−40 cm2 , computed with vs. without AE reconstruction. Valence states included are the first
four bands below the band gap, and conduction states included are all bands up to Edft = 60 eV.
3.1 Valence to Conduction

We begin with valence to conduction band transitions. The initial (final) states are indexed
by band number, i(f ), and Bloch momentum, ki (kf ) inside the 1BZ. The wave functions
in Eq. (2.1) can be substituted into the crystal form factor in Eq. (3.6),
Z
1
fi,ki →f,kf = d3 x ei(ki −kf +q)·x u∗f,kf (x) ui,ki (x)
V
X Z
1
= δq, kf −ki +G d3 x eiG·x u∗f,kf (x) ui,ki (x) , (3.14)
Ω cell
G
where the integral is over the primitive cell with volume Ω, and we have used the identity
P iq·r P
re = N G δq,G . The total rate in Eq. (3.12) is then
Nv XNc Z
2πσ e ρχ X d3 ki d3 kf X fe 2 2
R= 2 3 3
Fmed g(q, ω)
µχe mχ ρT 1BZ (2π) (2π) fe0
i=1 f =1 G
– 14 –
Z 2
1

× 3
d xe uf,kf (x) ui,ki (x) ,
iG·x ∗
(3.15)
Ω cell
where q = kf − ki + G, Nv(c) is the number of valence (conduction) bands. This is
identical to the rate formulae derived in [10, 15, 16] but written in terms of the periodic
Bloch functions, ui,k (x), instead of their Fourier transformed components, u ei,k,G , similar
to Ref. [50]. Numerically the position space form is superior since the integral over the
primitive cell can be computed by Fast Fourier Transform. This reduces the computational
complexity from O(NG2 ) to O(NG log NG ), where NG is the number of G points, i.e. the
number of Fourier components in the expansion of u ei,k in Eq. (2.3).
In Fig. 5 we show the scattering rate from valence to conduction transitions binned in
energy deposition, defined in Eq. (3.13), for a 1 GeV DM. The main difference between the
calculation performed here and in previous works is the effect of the AE reconstruction,
as discussed in Sec. 2.1.1. For the case of DM with a heavy mediator, the rate, even with
experimental thresholds as low as ∼ 10 eV, is significantly enhanced relative to previous
work. The AE reconstruction plays less of a role in the light mediator case since the
transition rate is dominated by small momentum transfers. However, at high thresholds,
where only larger momentum components can contribute, the AE reconstruction can still
significantly boost the scattering rate by fully including the contributions neglected in the
pseudo wave functions.
Since most earlier works computing DM-electron scattering include only valence to
conduction transitions, it is useful to understand for which DM masses these are the only
kinematically allowed transitions. If ω < Eg − Emax core , where E core is the maximum energy
max
of the core states, then the core states cannot contribute; if ω < Edft the free states are not
core } only the valence to conduction transitions
available. Therefore if ω < min{Edft , Eg − Emax
are allowed, which can be related to a DM mass via ωmax (mχ ) < min{Edft , Eg − Emax core },
where
1 m v 2
2 χ max
ωmax (mχ ) = mχ vmax = 3.9 eV , (3.16)
2 MeV 840 km/s
core = −116 eV
with vmax = ve + vesc , the maximum incoming DM velocity. For Si (Ge), Emax
(−28 eV), this corresponds to
(
15.2 MeV (Si) ,
mχ < (3.17)
7.8 MeV (Ge) .
Requiring that ωmax > Eg , where Eg is the band gap, sets a lower bound on the minimum
detectable mass, mmin
χ ,

2Eg Eg 840 km/s 2
mmin
χ = 2
= 0.25 MeV . (3.18)
vmax eV vmax
For Si (Ge), with a band gap of 1.11 (0.67) eV, mmin
χ is 0.28 (0.17) MeV. Lastly, we remark
that for DM interactions characterized by higher-dimensional operators (not considered in
this work), the scattering rate scales with higher powers of q and therefore is even more
sensitive to AE reconstruction (and also c → c contributions discussed below in Sec. 3.3),
which must be included in the analysis.
– 15 –
1 1
10−1 Fmed
10−1 Fmed
10−2 − d log
d log q = 2 10−2 − d log
d log q = 2
10−3 10−3
v→c
∆Rω × kg · yr
∆Rω × kg · yr
10−4 10−4
10−5 10−5 v→c
10−6 10−6
10−7 10−7
10−8 v→f 10−8 v→f
10−9 10−9
10−10 10−10
10−11 Si 10−11 Ge
0 25 50 75 100 125 150 0 25 50 75 100 125 150
ω [eV] ω [eV]
1 1
Fmed Fmed
10−1 − d log
d log q = 0 10−1 − d log
d log q = 0
v→c
v→c
∆Rω × kg · yr
10−2 ∆Rω × kg · yr
10−2
10−3 10−3
10−4 v→f 10−4 v→f
10−5 10−5
Si Ge
0 25 50 75 100 125 150 0 25 50 75 100 125 150
ω [eV] ω [eV]
Figure 6. DM-electron scattering rate from valence to conduction (v→c) bands and from valence
bands to free states (v→f) binned in energy deposition (with ∆ω = 1 eV) for 1 GeV DM, light (top
row) and heavy (bottom row) mediators, assuming σ e = 10−40 cm2 . The upper edge of the shaded
region corresponds to using Zeff from Eq. (2.15), while the bottom edge corresponds to Zeff = 1.
3.2 Valence to Free

For valence to free transitions the initial states are identical to those from Sec. 3.1, labeled
by band number i and Bloch momentum, ki . The final state wave functions are simple
plane waves given by Eq. (2.12), labeled by a momentum kf in the 1BZ with the bands
labeled by G. We can therefore directly substitute Eq. (2.13) into Eq. (3.14) derived in the
previous subsection, and obtain the crystal form factor:
X Z
1
fi,ki →Gf ,kf = δq, kf −ki +G d3 x ei(G−Gf )·x ui,ki (x)
Ω cell
G
X
= ei,ki ,Gf −G ,
δq, kf −ki +G u (3.19)
G
ei,ki ,G are the Fourier components of the Bloch wave functions defined in Eq. (2.3).
where u
Incorporating the Fermi factor correction discussed in Sec. 2.3, we find the rate in Eq. (3.12)
– 16 –
is given by
Nv X Z
2πσ e ρχ X d3 ki d3 kf X fe 2 2
R= 2 3 3
F (νi,ki ) 0
2
Fmed g(q, ω) uei,ki ,Gf −G ,
µχe mχ ρT 1BZ (2π) (2π)
i=1 Gf
fe
G
(3.20)
where
|kf + Gf |2 i,ki
ω≡ − Ei,ki , νi,ki = ν(Zeff , ω + Ei,ki ) . (3.21)
2me
With a change of variables, G0 = Gf − G and defining k0 ≡ kf + Gf (and then dropping
the prime for simplicity), the rate becomes
Nv Z Z 2
2πσ e ρχ X d3 ki X 2 d3 k fe 2
R= 2 3
F (νi,ki ) |e
ui,ki ,G | 3
Fmed g(q, ω) . (3.22)
µχe mχ ρT 1BZ (2π) (2π) fe0
i=1 G
where q = k − ki − G.
In Fig. 6 we compare the binned rate from the valence to conduction (v→c) calculation
in the previous subsection to the valence to free (v→f) one performed here, again for a
1 GeV DM. We see that for large ω, where the v→c calculation is limited by the number
of conduction bands included, the v→f calculation extrapolates the results to higher ω as
expected. There is some uncertainty due to the choice of the effective charge parameters,
which is why the results are shown in bands. The lower edge corresponds to the conservative
i,ki
choice of Zeff = 1 for all i, ki , and the upper edge corresponds to the value set by the binding
i,ki
energy, Eq. (2.15) with EB = −Ei,ki . We find that the conservative choice Zeff = 1 is a
better match to the edge for the v→c calculation, and will use this in our final projections
in Sec. 4. Note that as the threshold increases, the effect of v→f transitions becomes more
important, and for a heavy mediator non-negligible constraints can be placed even with
O(100) eV energy thresholds.
3.3 Core to Conduction

We now turn to core to conduction transitions. The initial core states are indexed by κ,
the atom in the primitive cell, the usual atomic quantum numbers, n, l, m, and the Bloch
momentum, ki . The final states are the DFT computed conduction states. The crystal
form factor is simply obtained from Eq. (3.14) by substituting ui,ki → uκnlm,ki :
X Z
1
fκnlm,ki →f,kf = δq, kf −ki +G d3 x eiG·x u∗f,kf (x) uκnlm,ki (x) , (3.23)
Ω cell
G
The total scattering rate is then
Na XNpκ n−1 l Nc Z
2πσ e ρχ X X X X d3 ki d3 kf X fe 2 2
R= 2 Fmed g(q, ω)
µχe mχ ρT (2π)3 (2π)3 fe0
κ=1 n=1 l=0 m=−l f =1 1BZ G
Z 2
1
× 3
d xeiG·x
uf kf (x) uκnlmki (x) ,
∗
(3.24)
Ω cell
– 17 –
1 1
10−1 Fmed
10−1 Fmed
10−2 − d log
d log q = 2 10−2 − d log
d log q = 2
10−3 10−3
3d
∆Rω × kg · yr
∆Rω × kg · yr
10−4 10−4
10−5 v→c 10−5
10−6 10−6
10−7 2p 10−7
10−8 10−8 3p
2s v→c
10−9 10−9
10−10 10−10 3s
10−11 Si 10−11 Ge
0 50 100 150 200 250 300 0 50 100 150 200 250 300
ω [eV] ω [eV]
10 10
1 Fmed 1 3d Fmed
− d log
d log q = 0 − d log
d log q = 0
10 −1 v→c 10 −1
∆Rω × kg · yr
∆Rω × kg · yr
10−2 2p 10−2
3p
10−3 2s 10−3
10−4 10−4 v→c 3s
10−5 10−5
10−6 10−6
Si Ge
0 50 100 150 200 250 300 0 50 100 150 200 250 300
ω [eV] ω [eV]
Figure 7. DM-electron scattering rate from core states to conduction bands binned in energy
deposition (with ∆ω = 5 eV) for 1 GeV DM, light (top row) and heavy (bottom row) mediators,
assuming σ e = 10−40 cm2 . The core states are labelled by the corresponding atomic orbitals, and
the conduction states up to Edft = 60 eV are included. For comparison we also show the v→c
contribution (after AE reconstruction) from Fig. 5 in gray.
where Na is the number of atoms in the primitive cell, Npκ is the largest principal quantum
number for atom κ, and ω = Ef,kf − Eκnl . The core wave functions uκnlm,ki (x) are given
by Eq. (2.9), and involves a sum over primitive cells. Since the integral in Eq. (3.23) is just
over one primitive cell, only the atoms in this and neighboring cells can have a significant
contribution. In other words, the sum over r converges very quickly due to the localized
nature of atomic wave functions. We therefore restrict r to be summed over only the 3×3×3
nearest cells.
The contribution of core to conduction (c→c) transitions to the binned rate, for mχ =
1 GeV, can be seen in Fig. 7. In most cases the v→c transitions are dominant compared
to the c→c, but there are two main scenarios where this is not true. First, when the
experimental threshold is raised; this excludes the v→c transitions and causes the c→c
contribution to be dominant. For example, consider a Si detector and a DM model with a
– 18 –
1 10
10 −1 AE Fmed Fmed
no AE − d log
d log q = 2
1 − d log
d log q = 0
10−2
core
10−3 10−1
∆Rω × kg · yr
∆Rω × kg · yr
10−4 10−2
10−5
10−6 10−3
10−7 10−4
10−8
10−5
10−9
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140
ω [eV] ω [eV]
Figure 8. Contribution to the DM-electron scattering rate binned in energy deposition (with
∆ω = 1 eV) from 3d electrons to conduction bands in Ge, for 1 GeV DM, light (left) and heavy
(right) mediators, assuming σ e = 10−40 cm2 . The three curves in each panel are computed using
DFT with and without AE reconstruction, and using the semi-analytic core wave functions.
heavy mediator (bottom left panel of Fig. 7). If the experimental threshold is ∼ 50 eV the
c→c contribution from the 2p states in Si gives the dominant contribution. Second, for a
Ge target, and a DM model with a heavy mediator, the 3d states dominate the rate even
at the lowest experimental threshold. To understand this in more detail we present Fig. 8
which compares the binned rate taking different modeling approaches for the 3d states in
Ge. We see that the large momentum components of the wave function, recovered only after
AE reconstruction in the DFT calculation, dominate the rate, which explains why previous
works have underestimated the importance of 3d electrons. Meanwhile, we see explicitly
at the scattering rate level that the semi-analytic approach accurately reproduces the DFT
calculation at low ω, and extends the latter beyond its cutoff at high ω, consistent with the
observation at the wave function level in Fig. 3.
3.4 Core to Free

The last transition type we consider involves a core electron initial state and a free electron
final state. The crystal form factor is most easily obtained by substituting Eqs. (2.8) and
(2.12) into its definition, Eq. (3.6):
Z
1 X iki ·(r+xκ )
fκnlm,ki →Gf ,kf = √ e d3 x ei(q−kf −Gf )·x ψκnlm
atom
(x − r − xκ )
NV r
X Z
1 i(ki +q−kf −Gf )·xκ
=√ e e i(q−kf +ki )·r
d3 x ei(q−kf −Gf )·x ψκnlm
atom
(x)
NV r
1 X
= √ ei(ki +q−kf −Gf )·xκ δq−kf +ki ,G ψeκnlm
atom
(−ki + G − Gf ) , (3.25)
Ω G
– 19 –
10−5 10−2
10−6 Fmed
10−3 Fmed
2p − d log
d log q = 2 10−4 3d − d log
d log q = 2
10−7
10−8 10−5
∆Rω × kg · yr
∆Rω × kg · yr
2s 10−6
10−9
c→c 10−7
10−10 3p
c→f 10−8
10−11 3s
10−9
10−12 10−10
10−13 Si 10−11 Ge
10−14 10−12
0 200 400 600 800 1000 0 200 400 600 800 1000
ω [eV] ω [eV]
10−1 10
2p − d log Fmed
1 3d − d log Fmed
10−2 d log q = 0 d log q = 0
2s 10−1
∆Rω × kg · yr
10−3 ∆Rω × kg · yr
10−2 3p
10−4
10−3 3s
−5
10
10−4
10−6 Si 10−5 Ge
0 200 400 600 800 1000 0 200 400 600 800 1000
ω [eV] ω [eV]
Figure 9. DM-electron scattering rate from core states to conduction bands (c→c) and to free
states (c→f) binned in energy deposition (with ∆ω = 10 eV) for 1 GeV DM, light (top row) and
heavy (bottom row) mediators, assuming σ e = 10−40 cm2 . As in the v→f calculation in Fig. 6, the
upper edge of the shaded bands corresponds to Zeff from Eq. (2.15), and the lower edge corresponds
to Zeff = 1.
where the Fourier transform of the RHF Slater type orbital (STO) core wave functions,
given in Eq. (2.10), are known analytically [71]:
Z
ψeSTO (q; Z, n, l, m) = d3 x eiq·x RSTO (x; Z, n) Ylm (x̂) ≡ χSTO (q; Z, n, l) Ylm (q̂) , (3.26)
l b(n−l)/2c
X
n ia0 q ωsnl
χSTO (q; Z, n) = 4πN (n − l)!(2Z) , (3.27)
Z
s=0 ((a0 q)2 + Z 2 )n−s+1
−s (n − s)!
ωsnl = −4Z 2 . (3.28)
s!(n − l − 2s)!
The direct detection rate is then
Npκ n−1
Na X l Z
2πσ e ρχ X X X d3 ki d3 kf
R= 2 3 3
µχe mχ ρT Ω 1BZ (2π) (2π)
κ=1 n=1 l=0 m=−l
– 20 –
10 11 10
13
3
Si Ge
10 −
10 −
10 12
9
q×
q×
11
8 10
=
ω
ω
ω = Eg 7 9
ω [eV]
ω [eV]
8
1 6 1 ω = Eg 7
5 6
4 5
4
3
3
2 2
10−1 1 10−1 1
1 10 1 10
q [keV] q [keV]
Figure 10. Dielectric function (q, ω), given by Eq. (3.31) with the parameters in Table 1, of Si
(left) and Ge (right) used to incorporate screening effects. The solid line indicates the edge of the
kinematically accessible region ω . qv. The dashed line is the band gap of the target. While the
static dielectric can be O(10), in the kinematically allowed region (q, ω) is an O(1) number, leading
to an O(1) effect on the scattering rates when the latter are dominated by small q, ω transitions.
XX 2 2
fe atom
× F (νκnl ) 2
Fmed g(q, ω) ψeκnlm (−ki + G − Gf ) , (3.29)
fe0
Gf G
κnl , ω + E
where q = kf − ki + G, and νκnl = ν(Zeff κnl ). We can now shift the Gf variable,
G ≡ Gf − G and define k = ki + G and k0 = kf + G. Therefore, q = k0 − k and
0 0
Npκ n−1
Na X l Z 2 2
2πσ e ρχ X X X d3 k d3 k 0 fe 2 eatom
R= 2 F (νκnl ) Fmed g(q, ω) ψκnlm (k) ,
µχe mχ ρT Ω (2π)3 (2π)3 fe0
κ=1 n=1 l=0 m=−l
(3.30)
which is the closest expression to the vacuum matrix element, with just the inclusion of the
core wave functions acting as a form factor.
In Fig. 9, we compare the binned rate from the core to conduction (c→c) calculation to
the core to free (c→f) calculation and see a reasonable extrapolation to higher ω. As with
the transition region between v→c and v→f shown in Fig. 6, we find Zeff = 1 gives a better
match between c→c and c→f. While the total number of electrons from these transitions is
expected to be much less than lower energy transitions, this is the best available calculation
for thresholds up to the kinematic limit of ωmax .
3.5 In-medium Screening

DM-electron interactions mediated by a dark photon or scalar are screened due to the in-
medium mixing between the mediator and the photon. The relevance of screening has been
recently emphasized in Ref. [48]. The screening factor, fe /fe0 , is related to the longitudinal
dielectric, fe /fe0 = (q̂ · · q̂)−1 , where is the dielectric tensor. It can be computed from
– 21 –
Target 0 α ωp [eV] qTF [keV]
Si 11.3 1.563 16.6 4.13
Ge 14 1.563 15.2 3.99
Table 1. Parameters used in the model of dielectric function, Eq. (3.31), of Si and Ge from Ref. [72],
which accounts for in-medium screening effects on the transition rate.
in-medium loop diagrams or extracted from optical data. Here we model the dielectric of
Si and Ge following Ref. [72]:
" 2 #−1
1 q 2 q4 ω
(q, ω) = 1 + +α + 2 2
− , (3.31)
0 − 1 qTF 4me ωp ωp
and ij = (q, ω) δij . Here, 0 ≡ (0, 0) is the static dielectric, α is a fitting parameter, qTF
is the Thomas-Fermi momentum, and ωp is the plasma frequency. The parameters used for
Si and Ge are listed in Table 1, and we plot the dielectric as a function of q, ω in Fig. 10.
Naively one might expect that the effect of the dielectric is to screen the rate by an
O(100) factor due to the fact that the static dielectric, 0 , is O(10). However, this is only
the value of the dielectric function at q = ω = 0, while as q → ∞ and ω → ∞ the dielectric
approaches unity. Therefore, the effect of the dielectric crucially depends on the region of
the kinematic phase space being probed. For a given energy deposition, ω, the momentum
transfer is limited to q & ω/v where v ∼ 10−3 is the DM velocity. Therefore, the absolute
minimum momentum transfer is qmin ∼ Eg /v ∼ O(keV), for O(eV) band gap targets. This
is parametrically the same size as the Thomas-Fermi momentum qTF , so the dielectric is
expected to slightly deviate from one, which causes only an O(1) shift to the scattering
rate, as seen in Fig. 11.
4 Projected Sensitivity
We now compile the results from the previous sections to compute the projected sensitivity.
We also compare the relative importance of each transition type, and discuss differences
between our results and previous calculations in the literature. When there are large dif-
ferences, it is typically because of the inclusion of AE reconstruction and core states in the
calculation. Since AE reconstruction and core states contribute predominantly at higher
momentum transfer and energy deposition, we will find the largest differences typically oc-
cur for a massive mediator and higher detector threshold, where the effects in some cases
can be more than an order of magnitude (especially for Ge). For the case of a massless
mediator and lower detection threshold, the differences with previous literature are much
smaller and mostly due to the inclusion of in-medium effects.
In Fig. 12 we show the contribution to the binned rate from each of the four transition
types, for a 1 GeV DM. We see that valence to conduction (v→c) has a higher peak than
the other three transition types, except for the Ge, heavy mediator case, where core to con-
duction (c→c) has the highest peak. For comparison, Refs. [10, 11] compute the valence to
– 22 –
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
∆Rωscr/∆Rωno scr
∆Rωscr/∆Rωno scr
0.7 Fmed
0.7
0.6 − d log
d log q = 2 0.6
Fmed
0.5 − d log
d log q = 0 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 Si 0.1 Ge
10 20 30 40 10 20 30 40
ω [eV] ω [eV]
1.1 1.1
1.0 1.0
0.9 0.9
0.8 0.8
0.7 0.7
Rscr/Rno scr
Rscr/Rno scr
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 Si 0.1 Ge
2 3 4 2 3
1 10 10 10 10 1 10 10 10 104
mχ [MeV] mχ [MeV]
Figure 11. Effect of screening on the binned rate (top row, for 1 GeV DM) and total rate (bottom
row, as a function of mχ ) from v→c transitions for DM models with a light (red) and heavy (blue)
mediator. The unscreened rate Rno scr is obtained with = 1, and the screened rate Rscr is obtained
with the model of the dielectric function given in Eq. (3.31).
conduction rates with DFT, including also the 3d states in Ge, but without AE reconstruc-
tion. As expected, we find a lower rate at the lowest energy depositions due to the inclusion
of in-medium screening, and a much higher rate at high ω due to AE reconstruction and
inclusion of core states.
The impact of these observations on the reach depends on the energy threshold. As-
suming charge readout (e.g. via a CCD), the relevant quantity is the number of electron-hole
pairs, Q, produced in an event. For an energy deposition ω, this is given by

ω − Eg
Q=1+ , (4.1)
ε
– 23 –
1 1
10−1 Fmed
10−1 Fmed
10−2 − d log
d log q = 2 10−2 − d log
d log q = 2
10−3 10−3
∆Rω × kg · yr
∆Rω × kg · yr
10−4 v→c 10−4
10−5 c→c 10−5
10−6 v→f 10−6
10−7 c→f 10−7
Derenzo et al.
10−8 10−8
10−9 10−9
10−10 10−10
10−11 Si 10−11 Ge
0 25 50 75 100 125 150 175 200 0 25 50 75 100 125 150 175 200
ω [eV] ω [eV]
1 1
Fmed Fmed
10−1 − d log
d log q = 0 10−1 − d log
d log q = 0
∆Rω × kg · yr
10−2 ∆Rω × kg · yr
10−2
10−3 10−3
10−4 10−4
10−5 10−5
Si Ge
0 25 50 75 100 125 150 175 200 0 25 50 75 100 125 150 175 200
ω [eV] ω [eV]
Figure 12. DM-electron scattering rate binned in energy deposition (with ∆ω = 1 eV) for 1 GeV
DM, light (top row) and heavy (bottom row) mediators, from all four transition types: valence to
conduction (v→c), valence to free (v→f), core to conduction (c→c), and core to free (c→f). We
assume σ e = 10−40 cm2 , and take Zeff = 1 for all effective charges in the Fermi factor. Note that
the c→c and c→f transitions involve semi-analytic treatment of 2p (3d) states and below in Si (Ge),
which has been validated with DFT calculations including AE reconstruction; see Fig. 3. We also
overlay the binned rate from Ref. [11] which computed the v→c contribution using QEdark (treating
3d states in Ge as valence, without including AE reconstruction effects).
where the values for ε are 3.6 eV and 2.9 eV for Si and Ge respectively. In Fig. 13, we show
the total rate as a function of the DM mass, for Q ≥ 1, 5, 10. The threshold only affects
the v→ c rate, as the other three transition types involve energy depositions corresponding
to Q > 10, and are therefore always fully included. We see that for Q ≥ 1, the valence
to conduction (v→c) contribution dominates the total rate with the exception of the Ge,
heavy mediator scenario, where core to conduction (c→c) is dominant for mχ & 30 MeV.
Higher thresholds significantly cut out v→c contributions in all cases, and render c→c more
important for Ge, even in the light mediator scenario. For Si, on the other hand, the total
rate is still dominated by v→c because the core states are much deeper and contribute a
– 24 –
102 102
Si Ge
1 1
R × kg · yr
R × kg · yr
10−2 10−2
10−4 10−4
10−6 10−6
Fmed Fmed
10−8 − d log
d log q = 2 10−8 − d log
d log q = 2
10−1 1 10 102 103 104 10−1 1 10 102 103 104

mχ [MeV] mχ [MeV]
102 102
Si Ge
1 1
R × kg · yr
R × kg · yr
10−2 10−2
10−4 v→c 10−4

v → c, Q ≥ 5
v → c, Q ≥ 10
10−6 v→f 10−6
c→c Fmed Fmed
10−8 c→f − d log
d log q = 0 10−8 − d log
d log q = 0
10−1 1 10 102 103 104 10−1 1 10 102 103 104

mχ [MeV] mχ [MeV]
Figure 13. DM-electron scattering rate as a function of the DM mass, for light (top row) and
heavy (bottom row) mediators, from all four transition types: valence to conduction (v→c), valence
to free (v→f), core to conduction (c→c), and core to free (c→f). We assume σ e = 10−40 cm2 , take
Zeff = 1 for all effective charges in the Fermi factor, and show results for several threshold Q values
which significantly impact the v→c contribution.
lower rate. We also see that v→f and c→f contributions are subdominant in all cases.
Finally, we present the projected reach on the DM-electron reference cross section σ e
in Figs. 14 and 15, for Q ≥ 1 and Q ≥ 10, respectively. Our new calculation yields several
important differences compared to the previous literature, and we discuss them in detail in
the following subsection.
4.1 Comparison With Previous Results

We begin by comparing to our previous work, Ref. [16], shown in brown in Fig. 14. We pre-
viously restricted our analysis to the light mediator scenario, and Q ≥ 1, which is relatively
– 25 –
10−38 10−38
Si Ge
10−39
10−39
10−40
σ e [cm2]
σ e [cm2]
−40
10
10−41
10−41
10−42
− d log Fmed
d log q = 2 − d log Fmed
d log q = 2
10−1 1 10 102 103 104 10−1 1 10 102 103 104

mχ [MeV] mχ [MeV]
10−38 10−38
This work
Si Griffin et al.
Essig et al.
Ge
10−39 Lee et al. 10−39
Knapen et al. no scr.
Knapen et al.
10−40 10−40
σ e [cm2]
σ e [cm2]
10−41 10−41
10−42 10−42
Q≥1 − d log Fmed
d log q =0 − d log Fmed
d log q = 0
10−1 1 10 102 103 104 10−1 1 10 102 103 104

mχ [MeV] mχ [MeV]
Figure 14. 95% C.L. exclusion reach (3 events) assuming 1 kg-year exposure, Q ≥ 1, for light (top
row) and heavy (bottom row) mediators. The results shown are from this work, Griffin et al. [16],
Essig et al. [10], Lee et al. [3], and Knapen et al. [48] (with and without screening). See Sec. 4.1 for
detailed comparison.
unaffected by AE reconstruction effects since the rate is peaked at small energy/momentum

transfers, as seen in Fig. 5. The main reason the reach here is weaker is the inclusion of
in-medium screening discussed in Sec. 3.5.
Next we compare to Ref. [10], shown in red in Figs. 14 and 15. Those results were
computed solely from valence to conduction (v→c) transitions. The largest discrepancy is
in the high mχ regime scattering off a Ge target via a heavy mediator. This is due to high
momentum contributions to the 3d wave functions in Ge. Ref. [10] computed the 3d states
with DFT without AE reconstruction, which as we saw in Fig. 3 is crucial for recovering
the dominant part of the 3d wave functions at high momentum. As discussed in Sec. 2.2,
our modeling of 3d electrons in Ge as core states reproduces their DFT-computed wave
functions up to the AE reconstruction cutoff, and provides a robust parameterization of
higher momentum components. Since the valence states in Ge also contribute an appre-
– 26 –
10−33 10−35
Si Ge
10−36
10−34
10−37
σ e [cm2]
σ e [cm2]
−35
10
10−38
10−36
10−39 This work
Essig et al.
Fmed Fmed
− d log
d log q =2 − d log
d log q = 2
10−1 1 10 102 103 104 10−1 1 10 102 103 104

mχ [MeV] mχ [MeV]
10−35 10−35
10−36 Si 10−36 Ge
10−37 10−37
σ e [cm2]
σ e [cm2]
10−38 10−38
10−39 10−39
10−40 10−40
10−41 Fmed 10−41 Fmed
Q ≥ 10 − d log
d log q = 0 − d log
d log q = 0
10−1 1 10 102 103 104 10−1 1 10 102 103 104

mχ [MeV] mχ [MeV]
Figure 15. 95% C.L. exclusion reach (3 events) assuming 1 kg-year exposure, Q ≥ 10, for light
(top row) and heavy (bottom row) mediators. The results shown are from this work and Essig et
al. [10]. See Sec. 4.1 for detailed comparison.
ciable amount, the Q ≥ 1 results in Fig. 14 only differ by about an order of magnitude.
However, the difference is more stark when going to higher Q thresholds in Fig. 15, which
essentially isolates the 3d electrons’ contribution. In the low mass regime the difference
is less significant, and primarily due to the inclusion of screening effects. Another differ-
ence that is important here is sampling of the 1BZ. Ref. [10] used a uniform 6 × 6 × 6
mesh with extra 27 points chosen by hand close to the center of the 1BZ, whereas here
(as well as in Ref. [16]) we use a uniform 10 × 10 × 10 grid. While checking convergence
we found our (unscreened) results using a 6 × 6 × 6 uniform mesh were a closer match to
Ref. [10]; generally, increasing the number of k points reduces the rate toward convergence,
i.e. R10×10×10 < R9×9×9 < R8×8×8 . This can be seen more directly in the difference be-
tween the brown and red lines in the light mediator scenario (as both are computed without
screening), and it affects Ge more than Si, as is expected due to the smaller band gap and
greater dispersions of nearby bands requiring denser k point sampling for convergence.
– 27 –
Ref. [3] also computed DM-electron scattering rates in semiconductors, focusing on Ge.
The approach taken in that paper was to semi-analytically model the Ge wave functions
with the core wave functions (with the same set of RHF STO wave function coefficients
tabulated in Ref. [43]) and treat the final states as free with a Fermi factor, analogous to
the core to free calculation performed here. As we can see from Fig. 14, while for most of
the mass range and mediators the estimates are too optimistic due to incorrect modeling of
the valence and conduction states, in the high mass region with a heavy mediator (bottom-
right panel), where 3d states dominate, their estimates are in good agreement with ours
presented here, as expected.
Finally, we discuss the comparison with the most recent work, Ref. [48], which was
limited to valence to conduction transitions. To show the effect of screening, we show
their projected reach with (purple) and without (green) screening in Fig. 14. Again the
largest discrepancy is in the heavy mediator scenario with a Ge target, primarily due to
the neglect of the 3d states in Ref. [48]. When these are not important, i.e. the low mass
regime or a light mediator, we generally find good agreement, with our reach being a bit
stronger. Notably this does not seem due to a mis-model of the dielectric, since the effect
of screening relative to our previous results, Ref. [16], is consistent with their result. We
also find that screening has a smaller effect at high masses in the heavy mediator scenario
for Si. These small differences are harder to disentangle since they could be due to: 1)
different xc-functionals used (PBE and HSE vs. TB09); 2) local field effects which are only
partially included here since we assume the screening factor is isotropic; 3) the plane wave
expansion parameter, Ecut , taken to be 500 eV without AE reconstruction in Ref. [48], vs.
1 keV, AE corrected to 2 keV taken here; 4) DM velocity distribution parameters, studied
in detail in Ref. [73], for which Ref. [48] assumed vesc = 500 km/s as opposed to vesc = 600
km/s chosen here; and 5) Ref. [48] took a directionally averaged dielectric, whereas here we
only assume isotropy in the screening factor but not the matrix element itself.
5 Conclusions
Dark matter-electron scattering in dielectric crystal targets, especially semiconductors like

Si and Ge, are at the forefront of DM direct detection experiments. It is therefore imperative
to have accurate theoretical predictions for the excitation rates. In this work, we extended
the scattering rate calculation in several key aspects. Much of the focus of previous cal-
culations has been on transitions from valence to conduction bands just across the band
gap, which will be accessible to near-future experiments. We performed state-of-the-art
DFT calculations for these states, and highlighted the importance of all-electron recon-
struction which has been neglected in most previous works. Along with this, we extended
the transition rate calculation by explicitly including the contributions from core electrons
and additional states more than 60 eV above the band gap using analytic approximations.
We updated the projected reach with our new calculation and found important differ-
ences compared to previous results. In particular, we found that in the heavy mediator
scenario, 3d electrons in Ge give a dominant contribution to the detection rate for DM
heavier than about 30 MeV. Also, the rate can be significantly higher than predicted previ-
– 28 –
ously for higher experimental thresholds. This is exciting because new DM parameter space
will be within reach even before detectors reach the single electron ionization threshold.
We also release a beta version of EXCEED-DM (available here [53]) that implements our
DM-electron scattering calculation for general crystal targets, and make the electronic wave
function data for Si and Ge [54], as well as the EXCEED-DM output [55], publicly available
so our present analysis can be reproduced. We have previously used EXCEED-DM for a tar-
get comparison study [16], and to study the daily modulation signals that can arise in
anisotropic materials [15]. The generality of EXCEED-DM means that potential applications
are vast. It can be used to compute detection rates for other target materials (assuming
DFT calculations of valence and conduction states are available), and can also be adapted
to include additional DM interactions such as in an effective field theory framework similar
to the study of atomic ionizations in Ref. [5] (see Ref. [74] for a recent effort in this direc-
tion). For momentum-suppressed effective operators, a full calculation in our framework
is even more important, as the effects of all-electron reconstruction and core states (over-
looked in Ref. [74]) are generally amplified. Moreover, the differential information that can
be obtained from our program facilitates further studies including realistic backgrounds.
Details of EXCEED-DM and additional example calculations will be presented in an upcoming
publication.
Acknowledgments
We are grateful to Kyle Bystrom for assistance with pawpyseed, and thank Alex Ganose,
Thomas Harrelson, Andrea Mitridate, Michele Papucci and Tien-Tien Yu for helpful discus-
sions. We also thank Rouven Essig and Adrian Soto for early correspondence about QEdark.
This material is based upon work supported by the U.S. Department of Energy, Office of
Science, Office of High Energy Physics, under Award No. DE-SC0021431 (TT, ZZ, KZ), by
a Simons Investigator Award (KZ) and the Quantum Information Science Enabled Discov-
ery (QuantISED) for High Energy Physics (KA2401032). This research used resources of
the National Energy Research Scientific Computing Center (NERSC), a U.S. Department
of Energy Office of Science User Facility located at Lawrence Berkeley National Laboratory,
operated under Contract No. DE-AC02-05CH11231. Some of the computations presented
here were conducted on the Caltech High Performance Cluster, partially supported by a
grant from the Gordon and Betty Moore Foundation. Work at the Molecular Foundry was
supported by the Office of Science, Office of Basic Energy Sciences, of the U.S. Department
of Energy under Contract No. DE-AC02-05CH11231.
References
[1] R. Essig, J. Mardon, and T. Volansky, “Direct Detection of Sub-GeV Dark Matter,” Phys.
Rev. D85 (2012) 076007, arXiv:1108.5383 [hep-ph].
[2] P. W. Graham, D. E. Kaplan, S. Rajendran, and M. T. Walters, “Semiconductor Probes of
Light Dark Matter,” Phys. Dark Univ. 1 (2012) 32–49, arXiv:1203.2531 [hep-ph].
– 29 –
[3] S. K. Lee, M. Lisanti, S. Mishra-Sharma, and B. R. Safdi, “Modulation Effects in Dark
Matter-Electron Scattering Experiments,” Phys. Rev. D 92 (2015) no. 8, 083517,
arXiv:1508.07361 [hep-ph].
[4] R. Essig, T. Volansky, and T.-T. Yu, “New constraints and prospects for sub-GeV dark
matter scattering off electrons in xenon,”Physical Review D 96 (aug, 2017) 043017,
arXiv:1703.00910 [hep-ph].
[5] R. Catena, T. Emken, N. A. Spaldin, and W. Tarantino, “Atomic responses to general dark
matter-electron interactions,” Phys. Rev. Res. 2 (2020) no. 3, 033195, arXiv:1912.08204
[hep-ph].
[6] DarkSide Collaboration, P. Agnes et al., “Constraints on Sub-GeV Dark-Matter–Electron
Scattering from the DarkSide-50 Experiment,” Phys. Rev. Lett. 121 (2018) no. 11, 111303,
arXiv:1802.06998 [astro-ph.CO].
[7] XENON Collaboration, E. Aprile et al., “Light Dark Matter Search with Ionization Signals
in XENON1T,” Phys. Rev. Lett. 123 (2019) no. 25, 251801, arXiv:1907.11485 [hep-ex].
[8] XENON Collaboration, E. Aprile et al., “Excess electronic recoil events in XENON1T,”
Phys. Rev. D 102 (2020) no. 7, 072004, arXiv:2006.09721 [hep-ex].
[9] R. Essig, A. Manalaysay, J. Mardon, P. Sorensen, and T. Volansky, “First Direct Detection
Limits on sub-GeV Dark Matter from XENON10,” Phys. Rev. Lett. 109 (2012) 021301,
[10] R. Essig, M. Fernandez-Serra, J. Mardon, A. Soto, T. Volansky, and T.-T. Yu, “Direct
Detection of sub-GeV Dark Matter with Semiconductor Targets,” JHEP 05 (2016) 046,
arXiv:1509.01598 [hep-ph].
[11] S. Derenzo, R. Essig, A. Massari, A. Soto, and T.-T. Yu, “Direct Detection of sub-GeV Dark
Matter with Scintillating Targets,” Phys. Rev. D 96 (2017) no. 1, 016026, arXiv:1607.01009
[hep-ph].
[12] Y. Hochberg, T. Lin, and K. M. Zurek, “Absorption of light dark matter in semiconductors,”
Phys. Rev. D95 (2017) no. 2, 023013, arXiv:1608.01994 [hep-ph].
[13] I. M. Bloch, R. Essig, K. Tobioka, T. Volansky, and T.-T. Yu, “Searching for Dark
Absorption with Direct Detection Experiments,” JHEP 06 (2017) 087, arXiv:1608.02123
[hep-ph].
[14] N. A. Kurinsky, T. C. Yu, Y. Hochberg, and B. Cabrera, “Diamond Detectors for Direct
Detection of Sub-GeV Dark Matter,” arXiv:1901.07569 [hep-ex].
[15] T. Trickle, Z. Zhang, K. M. Zurek, K. Inzani, and S. Griffin, “Multi-Channel Direct
Detection of Light Dark Matter: Theoretical Framework,” JHEP 03 (2020) 036,
arXiv:1910.08092 [hep-ph].
[16] S. M. Griffin, K. Inzani, T. Trickle, Z. Zhang, and K. M. Zurek, “Multichannel direct
detection of light dark matter: Target comparison,” Phys. Rev. D 101 (2020) no. 5, 055004,
arXiv:1910.10716 [hep-ph].
[17] S. M. Griffin, Y. Hochberg, K. Inzani, N. Kurinsky, T. Lin, and T. Chin, “Silicon carbide
detectors for sub-GeV dark matter,” Phys. Rev. D 103 (2021) no. 7, 075002,
arXiv:2008.08560 [hep-ph].
[18] P. Du, D. Egana-Ugrinovic, R. Essig, and M. Sholapurkar, “Sources of Low-Energy Events in
Low-Threshold Dark Matter Detectors,” arXiv:2011.13939 [hep-ph].
– 30 –
[19] Y. Hochberg, Y. Zhao, and K. M. Zurek, “Superconducting Detectors for Superlight Dark
Matter,” Phys. Rev. Lett. 116 (2016) no. 1, 011301, arXiv:1504.07237 [hep-ph].
[20] Y. Hochberg, M. Pyle, Y. Zhao, and K. M. Zurek, “Detecting Superlight Dark Matter with
Fermi-Degenerate Materials,” JHEP 08 (2016) 057, arXiv:1512.04533 [hep-ph].
[21] Y. Hochberg, T. Lin, and K. M. Zurek, “Detecting Ultralight Bosonic Dark Matter via
Absorption in Superconductors,” Phys. Rev. D94 (2016) no. 1, 015019, arXiv:1604.06800
[hep-ph].
[22] Y. Hochberg, Y. Kahn, M. Lisanti, K. M. Zurek, A. G. Grushin, R. Ilan, S. M. Griffin, Z.-F.
Liu, S. F. Weber, and J. B. Neaton, “Detection of sub-MeV Dark Matter with
Three-Dimensional Dirac Materials,” Phys. Rev. D 97 (2018) no. 1, 015004,
arXiv:1708.08929 [hep-ph].
[23] A. Coskuner, A. Mitridate, A. Olivares, and K. M. Zurek, “Directional Dark Matter
Detection in Anisotropic Dirac Materials,” arXiv:1909.09170 [hep-ph].
[24] R. M. Geilhufe, F. Kahlhoefer, and M. W. Winkler, “Dirac Materials for Sub-MeV Dark
Matter Detection: New Targets and Improved Formalism,” arXiv:1910.02091 [hep-ph].
[25] K. Inzani, A. Faghaninia, and S. M. Griffin, “Prediction of tunable spin-orbit gapped
materials for dark matter detection,” Physical Review Research 3 (2021) no. 1, 013069.
[26] DAMIC Collaboration, J. R. T. de Mello Neto et al., “The DAMIC dark matter
experiment,” PoS ICRC2015 (2016) 1221, arXiv:1510.02126 [physics.ins-det].
[27] DAMIC Collaboration, A. Aguilar-Arevalo et al., “Constraints on Light Dark Matter
Particles Interacting with Electrons from DAMIC at SNOLAB,” Phys. Rev. Lett. 123 (2019)
no. 18, 181802, arXiv:1907.12628 [astro-ph.CO].
[28] DAMIC, DAMIC-M Collaboration, M. Settimo, “Search for low-mass dark matter with
the DAMIC experiment,” in 16th Rencontres du Vietnam: Theory meeting experiment:
Particle Astrophysics and Cosmology. 4, 2020. arXiv:2003.09497 [hep-ex].
[29] SENSEI Collaboration, J. Tiffenberg, M. Sofo-Haro, A. Drlica-Wagner, R. Essig,
Y. Guardincerri, S. Holland, T. Volansky, and T.-T. Yu, “Single-electron and single-photon
sensitivity with a silicon Skipper CCD,” Phys. Rev. Lett. 119 (2017) no. 13, 131802,
arXiv:1706.00028 [physics.ins-det].
[30] SENSEI Collaboration, M. Crisler, R. Essig, J. Estrada, G. Fernandez, J. Tiffenberg,
M. Sofo haro, T. Volansky, and T.-T. Yu, “SENSEI: First Direct-Detection Constraints on
sub-GeV Dark Matter from a Surface Run,” Phys. Rev. Lett. 121 (2018) no. 6, 061803,
arXiv:1804.00088 [hep-ex].
[31] SENSEI Collaboration, O. Abramoff et al., “SENSEI: Direct-Detection Constraints on
Sub-GeV Dark Matter from a Shallow Underground Run Using a Prototype Skipper-CCD,”
Phys. Rev. Lett. 122 (2019) no. 16, 161801, arXiv:1901.10478 [hep-ex].
[32] SENSEI Collaboration, L. Barak et al., “SENSEI: Direct-Detection Results on sub-GeV
Dark Matter from a New Skipper-CCD,” Phys. Rev. Lett. 125 (2020) no. 17, 171802,
[33] SuperCDMS Collaboration, R. Agnese et al., “Search for Low-Mass Weakly Interacting
Massive Particles with SuperCDMS,” Phys. Rev. Lett. 112 (2014) no. 24, 241302,
arXiv:1402.7137 [hep-ex].
– 31 –
[34] SuperCDMS Collaboration, R. Agnese et al., “New Results from the Search for Low-Mass
Weakly Interacting Massive Particles with the CDMS Low Ionization Threshold
Experiment,” Phys. Rev. Lett. 116 (2016) no. 7, 071301, arXiv:1509.02448 [astro-ph.CO].
[35] SuperCDMS Collaboration, R. Agnese et al., “Projected Sensitivity of the SuperCDMS
SNOLAB experiment,”Phys. Rev. D95 (Apr, 2017) 082002, arXiv:1610.00006
[physics.ins-det]. https://link.aps.org/doi/10.1103/PhysRevD.95.082002.
[36] SuperCDMS Collaboration, R. Agnese et al., “Low-mass dark matter search with
CDMSlite,” Phys. Rev. D 97 (2018) no. 2, 022002, arXiv:1707.01632 [astro-ph.CO].
[37] SuperCDMS Collaboration, R. Agnese et al., “First dark matter constraints from a
SuperCDMS single-charge sensitive detector,”Physical Review Letters 121 (aug, 2018)
051301, arXiv:1804.10697 [hep-ex]. [Erratum: Phys.Rev.Lett. 122, 069901 (2019)].
[38] SuperCDMS Collaboration, R. Agnese et al., “Search for Low-Mass Dark Matter with
CDMSlite Using a Profile Likelihood Fit,” Phys. Rev. D 99 (2019) no. 6, 062001,
[39] SuperCDMS Collaboration, D. W. Amaral et al., “Constraints on low-mass, relic dark
matter candidates from a surface-operated SuperCDMS single-charge sensitive detector,”
Phys. Rev. D 102 (2020) no. 9, 091101, arXiv:2005.14067 [hep-ex].
[40] EDELWEISS Collaboration, E. Armengaud et al., “Searches for electron interactions
induced by new physics in the EDELWEISS-III Germanium bolometers,” Phys. Rev. D 98
(2018) no. 8, 082004, arXiv:1808.02340 [hep-ex].
[41] EDELWEISS Collaboration, E. Armengaud et al., “Searching for low-mass dark matter
particles with a massive Ge bolometer operated above-ground,” Phys. Rev. D 99 (2019)
no. 8, 082003, arXiv:1901.03588 [astro-ph.GA].
[42] EDELWEISS Collaboration, Q. Arnaud et al., “First germanium-based constraints on
sub-MeV Dark Matter with the EDELWEISS experiment,” Phys. Rev. Lett. 125 (2020)
no. 14, 141301, arXiv:2003.01046 [astro-ph.GA].
[43] C. Bunge, J. Barrientos, and A. Bunge, “Roothaan-Hartree-Fock Ground-State Atomic Wave
Functions: Slater-Type Orbital Expansions and Expectation Values for Z = 2-54,” Atom.
Data Nucl. Data Tabl. 53 (1993) 113–162.
[44] P. Giannozzi et al., “Quantum espresso: a modular and open-source software project for
quantum simulations of materials,” Journal of Physics: Condensed Matter 21 (2009) no. 39,
395502 (19pp). http://www.quantum-espresso.org.
[45] P. Giannozzi et al., “Advanced capabilities for materials modelling with quantum espresso,”
Journal of Physics: Condensed Matter 29 (2017) no. 46, 465901.
http://stacks.iop.org/0953-8984/29/i=46/a=465901.
[46] P. Giannozzi et al., “Quantum espresso toward the exascale,” The Journal of Chemical
Physics 152 (2020) no. 15, 154105, https://doi.org/10.1063/5.0005082.
https://doi.org/10.1063/5.0005082.
[47] Y. Hochberg, Y. Kahn, N. Kurinsky, B. V. Lehmann, T. C. Yu, and K. K. Berggren,
“Determining Dark Matter-Electron Scattering Rates from the Dielectric Function,”
arXiv:2101.08263 [hep-ph].
[48] S. Knapen, J. Kozaczuk, and T. Lin, “Dark matter-electron scattering in dielectrics,”
arXiv:2101.08275 [hep-ph].
– 32 –
[49] S. Knapen, J. Kozaczuk, and T. Lin, “DarkELF: A python package for dark matter
scattering in dielectric targets,” arXiv:2104.12786 [hep-ph].
[50] Z.-L. Liang, L. Zhang, P. Zhang, and F. Zheng, “The wavefunction reconstruction effects in
calculation of DM-induced electronic transition in semiconductor targets,” JHEP 01 (2019)
149, arXiv:1810.13394 [cond-mat.mtrl-sci].
[51] S. Griffin, S. Knapen, T. Lin, and K. M. Zurek, “Directional Detection of Light Dark Matter
with Polar Materials,” Phys. Rev. D 98 (2018) no. 11, 115034, arXiv:1807.10291 [hep-ph].
[52] A. Coskuner, T. Trickle, Z. Zhang, and K. M. Zurek, “Directional Detectability of Dark
Matter With Single Phonon Excitations: Target Comparison,” arXiv:2102.09567
[hep-ph].
[53] T. Trickle, “tanner-trickle/EXCEED-DM: EXCEED-DM-v0.1.0,” Zenodo, May, 2021.
https://doi.org/10.5281/zenodo.4747696.
[54] T. Trickle, “EXCEED-DM: DFT-computed electronic wave functions for Si and Ge,”
Zenodo, 2021.
[55] T. Trickle, “EXCEED-DM: Calculated Scattering Rate Data for Si and Ge,” Zenodo, 2021.
[56] P. Hohenberg and W. Kohn, “Inhomogeneous Electron Gas,” Phys. Rev. 136 (1964)
B864–B871.
[57] R. M. Martin, Electronic structure: basic theory and practical methods. Cambridge university
press, 2020.
[58] W. Kohn and L. J. Sham, “Self-Consistent Equations Including Exchange and Correlation
Effects,” Phys. Rev. 140 (1965) A1133–A1138.
[59] J. P. Perdew, K. Burke, and M. Ernzerhof, “Generalized Gradient Approximation Made
Simple,” Physical Review Letters 77 (1996) no. 18, 3865–3868.
http://link.aps.org/doi/10.1103/PhysRevLett.77.3865.
[60] J. Heyd, G. E. Scuseria, and M. Ernzerhof, “Hybrid functionals based on a screened
Coulomb potential,” Journal of Chemical Physics 118 (2003) no. 18, 8207–8215.
[61] J. Heyd, G. E. Scuseria, and M. Ernzerhof, “Erratum: Hybrid functionals based on a
screened Coulomb potential (Journal of Chemical Physics (2003) 118 (8207)),” Journal of
Chemical Physics 124 (2006) no. 21, . https://doi.org/10.1063/1.2204597.
[62] P. Blöchl, “Projector augmented-wave method,” Physical Review B 50 (1994) no. 24,
17953–17979.
[63] G. Kresse and D. Joubert, “From ultrasoft pseudopotentials to the projector
augmented-wave method,” Physical Review B 59 (1999) 1758–1775.
[64] G. Kresse and J. Hafner, “Ab initio molecular dynamics for liquid metals,” Physical Review
B 47 (1993) 558–561.
[65] G. Kresse and J. Hafner, “Ab initio molecular-dynamics simulation of the
liquid-metal–amorphous-semiconductor transition in germanium,” Physical Review B 49
(1994) no. 20, 14251–14269. http://link.aps.org/doi/10.1103/PhysRevB.49.14251.
[66] G. Kresse and J. Furthmüller, “Efficiency of ab initio total energy calculations for metals and
semiconductors using a plane-wave basis set,” Computational Materials Science 6 (1996)
15–50.
– 33 –
[67] G. Kresse and J. Furthmüller, “Efficient iterative schemes for ab initio total-energy
calculations using a plane-wave basis set,” Physical Review B 54 (1996) 11169–11186.
http://link.aps.org/doi/10.1103/PhysRevB.54.11169.
[68] C. Rostgaard, “The projector augmented-wave method,” arXiv:0910.1921
[cond-mat.mtrl-sci].
[69] K. Bystrom, D. Broberg, S. Dwaraknath, K. A. Persson, and M. Asta, “Pawpyseed:
Perturbation-extrapolation band shifting corrections for point defect calculations,”arXiv
e-prints (Apr, 2019) arXiv:1904.11572, arXiv:1904.11572 [cond-mat.mtrl-sci].
[70] G. H. S. E. U. Condon, The Theory of Atomic Spectra. Cambridge University Press,
January, 1935. https://www.ebook.de/de/product/4287067/e_u_condon_george_h_
shortley_the_theory_of_atomic_spectra.html.
[71] D. Belkić and H. S. Taylor, “A unified formula for the fourier transform of slater-type
orbitals,”Physica Scripta 39 (feb, 1989) 226–229.
[72] G. Cappellini, R. D. Sole, L. Reining, and F. Bechstedt, “Model dielectric function for
semiconductors,”Physical Review B 47 (apr, 1993) 9892–9895.
[73] A. Radick, A.-M. Taki, and T.-T. Yu, “Dependence of Dark Matter - Electron Scattering on
the Galactic Dark Matter Velocity Distribution,” JCAP 02 (2021) 004, arXiv:2011.02493
[hep-ph].
[74] R. Catena, T. Emken, M. Matas, N. A. Spaldin, and E. Urdshals, “Crystal responses to
general dark matter-electron interactions,” arXiv:2105.02233 [hep-ph].
– 34 –

Extended Calculation of Dark Matter-Electron Scattering in Crystal Targets

Uploaded by

Copyright:

Available Formats

Extended Calculation of Dark Matter-Electron Scattering in Crystal Targets

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Extended Calculation of Dark Matter-Electron Scattering in Crystal Targets

Uploaded by

Copyright:

Available Formats

CALT-TH-2021-018

Extended Calculation of Dark Matter-Electron

Sinéad M. Griffin,a Katherine Inzani,a Tanner Trickle,b Zhengkang Zhangb and

3 Electronic Transition Rates 12

Electronic excitations have been established as an alternative to nuclear recoils in direct

1.1 Overview of the Calculation and Key Results

2.1 DFT Wave Functions and Band Structures

In principle, DFT provides an exact solution to the many-electron Schrödinger equation by

2.1.1 All-electron Reconstruction

2.2 Atomic Wave Functions

where Ω = V /N is the primitive cell volume.

2.3 Plane Wave Approximation

uG,k (x) = eiG·x . (2.13)

(2π)3 Mσ0 s0 σs (p0 , k0 , p, k) (3) 0

3.1 Valence to Conduction

10−4 v→f 10−4 v→f

3.2 Valence to Free

3.3 Core to Conduction

The total scattering rate is then

3.4 Core to Free

3.5 In-medium Screening

10−1 1 10 102 103 104 10−1 1 10 102 103 104

10−4 v→c 10−4

10−1 1 10 102 103 104 10−1 1 10 102 103 104

4.1 Comparison With Previous Results

10−1 1 10 102 103 104 10−1 1 10 102 103 104

10−1 1 10 102 103 104 10−1 1 10 102 103 104

unaffected by AE reconstruction effects since the rate is peaked at small energy/momentum

10−1 1 10 102 103 104 10−1 1 10 102 103 104

10−1 1 10 102 103 104 10−1 1 10 102 103 104

Dark matter-electron scattering in dielectric crystal targets, especially semiconductors like

You might also like